Hashing Power! Is it all that matters?

[elabuwa]

Hi guys,

I am a bit new to the whole mining thing and just have a economical issue when it comes to selecting GPUs. Sorry if the title is in appropriate.

Let's just say in mining Ethereum, a 1070 ti (costs $945) has a speed of Approx 30-31 Mh/s, a 1080 ti (costs $1450) has a speed of approx 38 mh/s, a 1060 3gb (costs $350 used) has a speed of approx 20 mh/s.

Putting aside the brand new/used factor, PCI slot usage on the board, DAG size etc and considering only sheer economics($$$)/speed,

I could buy about 3 1060s and get a total speed of approx 60 mh/s rather than buy a 1070 ti. I could buy about 4 1060s a get a total speed of approx 80 mh/s.

Therefore, does buying 3 or4 1060s rather than 1 1080ti or 1070ti get me at an advantage in solving shares?

Does the speed/ram in a GPU has a part in playing the share solving scenario?
What would you recommend if I am at a budget in starting up this mining thingi?

I know individually the 1080ti will out perform the 1060s in it's sleep.

I just assumed that 1 GPU was meant to process 1 share/task.
If I have 6 GPUs 6 shares would be tried solved simultaneously. Was I wrong?
Do all GPU's combined together solve a share?

[pickleburglar]

It makes no difference how many cards you got or how fast they are individually. 2x30 is just as good as 1x60.

There is more than that when considering cards though

1. Bigger cards tend to be more power efficient per every hash calculated
2. Bigger cards take less pci-e slots than multiple small ones. With every new motherboard you also need cpu, ram and a drive so consider all their costs per pcie slot and add it to the price of the card.
3. Bigger cards tend to be more stable. From personal experience, when high end gpu's crash, they usually don't take the entire system down but give you a notification, recover and move on.. a luxury feature I guess.
4. Bigger cards are more "future proof". That 3GB 1060 is a great example - they only have around 1 year of hashing left on eth block chain, after that the DAG can't fit in its memory anymore. Sure they can do other algorithms but with the sudden shift of all 3GB 1060's to other algos/chains, their profits are going to be significantly reduced.


At the end of the day though, none of this matters, GPU's are in terribly short supply, you basically buy whatever you can get your hands on.

[Pekine]

Is there a specific reason you are looking for NVIDIA cards? I'm using AMD cards and they run pretty stable and are a lot cheaper than the prices you mention. I Buy my MSI AMD RX580 - 8GB at 300€ (Sapphire at 350€) and they run at a hashrate of 29 for the MSI and 32 for the Sapphire.

[ivakar]

If you are targeted to mine etherium then you'd better to buy amd rx card, fot example rx 570 can do 27-30mh and cost less then 500$

[elabuwa]

Is there a specific reason you are looking for NVIDIA cards? I'm using AMD cards and they run pretty stable and are a lot cheaper than the prices you mention. I Buy my MSI AMD RX580 - 8GB at 300€ (Sapphire at 350€) and they run at a hashrate of 29 for the MSI and 32 for the Sapphire.

There is no specific reason that I'm looking at NVIDIA.
But in order to to start mining with AMD, do we not need to flash the bios or what ever in that card which invalidates the warranty?

[elabuwa]

If you are targeted to mine etherium then you'd better to buy amd rx card, fot example rx 570 can do 27-30mh and cost less then 500$

I'm not targeting ETH.
Thinking of ZCash or may be ZClassic something in between 100-500 range.
Do we not need to change the bios etc of AMD cards?
With NVIDIA it's easy as running afterburner is what I have heard.

[chup]

No. It matters most how You sell Your mined coins.

[MarkAz]

Assuming you're doing this for more than one machine, it definitely does matter.  What everyone has been talking about is the raw profit per card - and what they've said is absolutely true, it's just a factor of how much return vs the price.  But if you're doing a larger deployment, then it gets a bit more complicated.

The first thing is that you also have overhead for the card - for instance, let's say you're using a MiningSky case, that holds 8 GPU's and costs $1000, that means you have a PER CARD overhead of $125 to be added to a card for what I typically call 'platform' ROI.  So let's say that a 1060 cost $250 and makes $2.5/day (to keep things simple) - you ROI's on the card in 100 days.  Now add the overhead of the case and your ROI is 150 days...  Now let's say you get a 1080 TI for $750 and it makes $7.5/day (once again, to keep things simple), you ROI in the same 100 days, but now the case ROI doesn't add 50 days, it only adds ~17 days.  This is why many times people like to buy the bigger cards...

Now here's the benefit of smaller cards - generally speaking, if you're going to run into problems typically it's about cooling or power, and they're both connected.  The 1080 TI can use 250w, while the 1060 is something like 100w - so your 8 card rig goes from 2000w of pure card consumption to 800w - which is a much easier number to deal with in terms of PSU cost as well as cooling.  Of course, the physical amount of space you'll need will be substantially larger, as in the example here it takes 3x as many 1060's to make one 1080 TI - but generally speaking space isn't an issue.

In terms of shares - this is just a factor of difficulty, and really doesn't matter much, in that it's just something the pool uses to prove that you're actually doing work.  If you were to run a local pool, you wouldn't even need to send partial shares, as you would implicitly trust all the devices.  Plus shares are just probabilistic - focus on hash rate for performance, and look at stale or bad shares as indications of problems in your network/hardware.

Hope that helps, and maybe makes you think about it in a larger context...